Microsoft killed the GPU mafia 🤯
They open-sourced bitnet.cpp, their 1-bit LLM inference framework. It can run a 100B-parameter BitNet model on a single CPU, no GPU needed.
- Up to 6.17x faster on x86 CPUs
- Up to 82.2% less energy
100% Open Source.
github.com/microsoft/BitNet
Repo: github.com/louis-e/arnis
Follow me for more practical AI tools like this every day.
4
307
i can't believe nobody caught this.
Anthropic's entire growth marketing team was ONE person
(for nearly 10 months, confirmed)
one non-technical marketer handled paid search, paid social, app stores, email and SEO for the company behind claude
here's how one person did the work of a full marketing team:
step 1 is a CSV.
1. he exports every existing ad from his ad platforms with their performance data (CTR, conversions, spend, etc)
2. drops the whole file into claude code
3. asks it to find what's underperforming
claude goes through the data, flags the weak ads, and writes new copy variations right there
here's the clever part:
he splits the job across 2 specialized sub-agents:
1. one that only writes headlines (max 30 characters)
2. one that only writes descriptions (max 90 characters)
each agent is tuned to its own constraint, so the output is far better than forcing both into one prompt
now he has hundreds of new headlines and descriptions.
but that's only the copy.
he still needs the actual creatives, the images and banners that run on facebook, google, etc.
so he built a figma plugin that:
1. grabs all the new headlines and descriptions
2. finds the ad templates in his figma files
3. swaps the copy into each one automatically
up to 100 ready-to-ship ad variations, at half a second per batch.
a job that used to mean hours of duplicating frames and pasting text by hand
now the ads are live.
next question: which ones actually work?
for that he built an MCP server (a custom integration that lets claude talk directly to outside tools) hooked into the meta ads API.
so he can ask claude things like:
• "which ads had the best conversion rate this week"
• "where am i wasting spend"
and get real answers from live campaign data without ever opening the meta ads dashboard
and the piece that closes the loop:
a memory system that logs every hypothesis and experiment result across ad iterations.
so when he goes back to step 1 for the next batch...
claude automatically pulls in what worked and what didn't from every previous round.
the system gets smarter every cycle.
that level of systematic testing across hundreds of ads would normally need a dedicated analytics person just to keep track
the numbers from anthropic's own doc:
ad creation went from 2 hours to 15 minutes. 10x more creative output.
and he's testing more variations across more channels than most full marketing teams
one of the most valuable AI companies in the world.
and their whole growth marketing operation (not GTM) = one person and claude code
wild
before anyone asks: yes, it's real.
confirmed by the one-man army himself ↓
teapawt.org/helloitsaustin/status/…
Replying to @itsolelehmann
when this was written i was the only person on growth marketing. i was a one person team for nearly 10 months. growth marketing != gtm
1
1
53
full doc from anthropic (see the growth marketing section): www-cdn.anthropic.com/58284b…
8
You can try it now on Google Labs: learnyourway.withgoogle.com/
Lessons on the immune system, economics, sociology and more. Each one adapts to your grade and interests.
Static textbooks had a good run. Personalized AI learning is here.
1
11



